The Politics of the United States' Bilateral Investment Treaty Program

Transcription

1 University of Chicago Law School Chicago Unbound Coase-Sandor Working Paper Series in Law and Economics Coase-Sandor Institute for Law and Economics 2015 The Politics of the United States' Bilateral Investment Treaty Program Adam S. Chilton Follow this and additional works at: Part of the Law Commons Recommended Citation Adam S. Chilton, "The Politics of the United States' Bilateral Investment Treaty Program" (Coase-Sandor Working Paper Series in Law and Economics No. 722, 2015). This Working Paper is brought to you for free and open access by the Coase-Sandor Institute for Law and Economics at Chicago Unbound. It has been accepted for inclusion in Coase-Sandor Working Paper Series in Law and Economics by an authorized administrator of Chicago Unbound. For more information, please contact

2 CHICAGO COASE-SANDOR INSTITUTE FOR LAW AND ECONOMICS WORKING PAPER NO. 722 (2D SERIES) The Politics of the United States Bilateral Investment Treaty Program Adam Chilton THE LAW SCHOOL THE UNIVERSITY OF CHICAGO March 2015 This paper can be downloaded without charge at: The University of Chicago, Institute for Law and Economics Working Paper Series Index: and at the Social Science Research Network Electronic Paper Collection. Electronic copy available at:

3 THE POLITICS OF THE UNITED STATES BILATERAL INVESTMENT TREATY PROGRAM Adam Chilton * March 10, 2015 Abstract Scholars consistently argue that the United States has signed Bilateral Investment Treaties (BITs) with developing countries to promote the development of international investment law and to protect American capital invested abroad. I challenge this view of the United States BITs program. I argue that the United States has used BITs as a foreign policy tool to improve relationships with strategically important countries in the developing world, and, as a result, the program should in part be evaluated based on whether it has produced political benefits. I empirically test this theory in two ways. First, I test whether investment or political considerations are better at explaining U.S. BIT signings. This analysis shows that investment considerations do not help to explain the pattern of U.S. BIT formation, but that political considerations do. Second, I estimate the political benefits the United States has received from signing BITs with developing states. This analysis suggests that having signed a BIT makes countries likely to vote similarly to the United States at the United Nations. This project thus provides the first empirical evidence that the U.S. BITs program has been motivated by political considerations, and that the program may have produced modest foreign policy dividends. * Assistant Professor of Law, University of Chicago Law School. I would like to thank Daniel Abebe, Rachel Brewster, Britt Cramer, Tom Ginsburg, Tom Miles, Rich Nielsen, Eric Posner, Rob Schub, Beth Simmons, Matthew Stephenson, Dustin Tingley, Jason Yackee, and Mark Wu for helpful comments and advice. I would also like to thank Daniel Marcin for helpful research assistance. This draft benefited from presentations at the American Law & Economics Association 2014 conference, the American Society of International Law 2014 conference, the Midwest Political Science Association 2014 conference, the ASIL International Economic Law Interest Group Junior Scholars Forum, the University of Chicago Law School Works in Progress Workshop, and the Harvard International Relations Research Workshop. 1 Electronic copy available at:

4 I. INTRODUCTION In the last sixty years, over 2,600 Bilateral Investment Treaties (BITs) have been negotiated between pairs of countries (Salacuse 2010, 428). Taken together, these treaties create a regime of international law that provides protections for individuals and corporations seeking to invest their capital abroad. Although the United States ranks first in the world in both foreign direct investment inflows and outflows (Sachs and Sauvant 2009), America was a late entrant into the BITs regime. The United States did not express a willingness to sign BITs until 1981 twenty-two years after Germany negotiated the world s first BIT and the United States did not have its first BIT in effect until 1988 (Salacuse 2010, 433). Despite this slow start, the United States has now signed BITs with 47 countries. 1 All U.S. BITs are based on a model treaty and, although the specific provisions of the model have evolved over the years, the core elements of these agreements have been the same (Akhtar and Weiss 2013, 8). First, agreements guarantee that investments made by individuals and corporations from the other country will be treated fairly and equitably. Second, the agreements limit expropriation of investment, and provide for compensation when expropriation does take place. Third, the agreements provide investors the right to transfer their property out of the foreign state freely. Fourth, the agreements place restrictions on trade-distorting performance requirements like local content requirements or export quotas. Fifth, if the terms of the BIT have been violated and the national courts of the foreign country do not provide redress, the agreements authorize investors to force the foreign state to participate in binding arbitration. Taken together, these five provisions give assurances to capital exporters that investments made in the market of a treaty partner will be provided with legal protection. Perhaps unsurprisingly, as the number of BITs that the United States is party to has proliferated, so has scholarship on these agreements (Shaffer and Ginsburg 2012, 35-38). Although this scholarship has not included a single empirical study evaluating the motivations of the U.S. BITs program, scholars have consistently argued that the United States signed BITs because it hoped to influence the development of international investment law and to protect American investments abroad (Elkins, Guzman, and Simmons 2006, ; Vandevelde 2005, 171; Lang 1998, 457; Vandevelde 1998a, 201-2; Vandevelde 1988,1-2; Gann 1985, 374; Sachs 1 Appendix A provides a complete list of the BITs the United States has signed. 2 Electronic copy available at:

5 1984, 195). Given this belief, scholars have evaluated U.S. BITs almost entirely based on their ability to protect and promote investment. Although it is certainly at least partially true that the United States was concerned with the development of international investment law and hoped that these treaties would help American investors, there are several reasons to doubt that this investment-centric view of the U.S. BITs program is the fully story. First, U.S. BIT negotiators have warned treaty partners that they should not expect a wave of new investments as a consequence of these agreements (Alvarez 2010, 621 n.69; Vandevelde 1998a, 212), which is evidence that U.S. officials themselves are aware that the economic impact of these agreements is likely quite limited. Second, if the United States were motivated by a desire to promote the development of international investment law and to protect American investors, it would have entered into a BIT with any country that would agree to its terms. This, however, was not the case (Vandevelde 1993, ). Third, there is scant evidence of any pressure from American interest groups on the United States to ratify the BITs that it has signed, suggesting that U.S. investors are not eager to avail themselves of any new opportunities or protections that BITs may provide. Fourth, evidence suggests that U.S. BITs do not have a positive impact on investment flows between the United States and partner countries (Peinhardt and Allee 2012; Yackee 2008). Fifth, there is reason to believe that BITs do not influence companies investment decisions (Yackee 2010), which calls into question whether BITs are negotiated to provide increased protections for capital exporters. Given the limited interest in and evidence of the investment benefits of BITs, it is worth reconsidering why the United States actively pursued a BITs program for three decades. I argue that the dominant narrative misses the mark: the United States did not primarily form BITs to promote and protect investments abroad, but instead to improve relationships with politically important developing countries. BITs have been used in this way because they have several features that make them an appealing foreign policy tool: they do not necessitate outlaying funds, require the United States to make only redundant obligations, are easy to sell politically, and take minimal effort to negotiate given their standardized nature. Having a BIT with the United States is attractive politically to the governments of many developing countries as well. Potential BIT partners are frequently eager to sign these treaties, even though they are warned that the agreements likely will not lead to new investments, because the treaty provides domestic 3

6 political benefits to the country s government. Since international relations are a repeated game, a new treaty partner would thus likely reciprocate by extending political benefits to the United States in order to receive favorable treatment again in the future. The implication of this theory is that investment considerations will not explain which countries the United States has chosen to sign BITs with, and that looking solely at the investment consequences of the treaties ignores many of their potential benefits. Instead, it is my contention that political considerations will better explain which countries the United States has chosen to sign BITs with, and that BITs should be evaluated at least in part based on whether they have generated political dividends. I empirically test this theory in two ways. First, I test what factors predict the countries that the United States has signed BITs with over the last thirty years. To do so, I use a series of survival models that estimate the likelihood that a BIT will be signed between the United States and a potential treaty partner in a given year. The results of this analysis suggest that variables that measure investment considerations specifically a country s level of investment risk and the amount of investment flows from the United States do not have a statistically significant impact on the likelihood that the United States will sign a BIT with that country. Moreover, the results of this analysis demonstrate that variables that measure the political importance of developing countries specifically whether the state received military aid or was formerly communist do have a positive and statistically significant impact on the likelihood that the country would sign a BIT with the United States. These results thus lend clear support to my theory that the United States selected BIT partners based on political, and not investment, considerations. Second, I examine whether signing BITs has produced political benefits for the United States. To do so, I use a recently developed statistical method life history matching that helps mitigate the problems of selection bias and post-treatment bias that have plagued most timeseries cross-sectional research designs (Nielsen, n.d.), and that has shown potential as a way to study the effects of treaty ratification (Nielsen and Simmons 2014). I specifically use this method to examine whether signing a BIT with the United States made the treaty partner more likely to: (1) vote similarly to the United States in the United Nations General Assembly; (2) allow the United States to deploy troops within its territory; or (3) support the invasion of Iraq in My results suggest that having a BIT in effect with the United States cause the treaty partner to be more supportive of the United States in the UN General Assembly, but likely did not alter the 4

7 number of U.S troop deployed on the partner s soil or whether the partner joined the Iraq War Coalition. In other words, this study not only shows that the United States chose which states to pursue BITs with based on political considerations, but also that signing BITs likely produced modest political dividends. This project thus makes several important contributions to the literature on international investment law. First, this project should cause commentators and scholars to change the narrative offered to explain why the United States has pursued BITs with developing states. The existing literature on U.S. BITs almost exclusively claims that these treaties were negotiated to produce more favorable conditions for American investors and promote the development of proinvestor international law. After this project, this account should be replaced with the new explanation that BITs have been negotiated as a low cost foreign policy tool. Second, this paper is the first quantitative effort to consider directly why the United States might have chosen to pursue BITs with specific countries. Previous efforts have focused on explaining the proliferation of BITs from developing countries perspectives (Alvarez 2010; Tobin and Busch 2010; Guzman 1998), but have not flipped the question and asked why the United States or developed countries more broadly would take the time to negotiate and ratify investment treaties that have limited direct influences on capital flows. Third, this project develops a theory that tries to explain both why the United States and developing countries would be motivated to sign investment treaties, even if they do not promote investment, and to explain why those treaties could still result in dividends albeit political ones for the United States. Finally, this project provides evidence suggesting that the United States BITs program has produced benefits that have not yet been understood or explored. This evidence should cause scholars who are skeptical of the economic benefits of BITs to reassess the program based on the political advantages that they may provide. This paper proceeds as follows. Part II describes conventional explanations for the growth of the United States BITs program, before developing a political theory of BITs formation. Part III tests the theory by empirically assessing the factors that predict the BITs the United States has chosen to sign, and Part IV tests the theory by empirically examining whether BITs have produced political dividends. Part V concludes. 5

8 II. A POLITICAL THEORY OF THE UNITED STATES BITS PROGRAM Although scholars have nearly universally argued that the United States BITs program was motivated by investment considerations, I argue that the United States primarily used BITs as a foreign policy tool. In order to make that argument, I first provide an overview of the United States history with BITs. Second, I lay out the standard explanation of the United States BITs program: namely, the idea that the program was motivated by a desire to protect American investments abroad. Third, I present a range of evidence suggesting that the United States BITs program cannot be explained purely as an effort to promote investment law and protect investors. Finally, I develop the theory that the United States used BITs as a low-cost method of cultivating relationships with politically important countries in the developing world. A. The United States Experience with BITs In 1977, twenty years after European countries first began to negotiate bilateral treaties that regulate investments, the United States launched its BITs program (Vandevele 1993). 2 The program was a successor to the treaties of general relations Friendship, Commerce, and Navigation (FCN) Treaties that included investment provisions. The United States had negotiated FCNs for nearly two hundred years, but that had ground to a halt by the mid-1960s. 3 After starting the BITs program, however, the United States spent the next four years preparing a model treaty that could be used in negotiations. This process was delayed in party due to extensive interagency consultations, and the transfer of primary authority over the program from the State Department to the Office of the United States Trade Representative. In 1981, the United States finally produced a draft model BIT to use in negotiations with other countries (Vandevelde 1993, 627). 2 There have already been a number of detailed accounts of the history of the United States BITs program, so I offer only a very brief account here. For a detailed discussion of the establishment and development of the U.S. BITs program, see Vandevelde (1993, ); Vandevelde (1992); Gann (1985, 373); Sachs (1984). 3 For a general discussion of the end of the FCN program, see Coyle (2013). Although a number of factors contributed to the end of the program, one driving factor was that many of the topics regulated by the agreements began to be governed by other sources of international law. 6

9 Figure 1: Countries That Have Signed a BIT with the United States Figure 1 shows the countries that have signed a BIT with the United States. Countries that signed a BIT that is currently in effect are solid, and countries that signed a BIT that is not currently in effect are dashed. In 1982, the United States started negotiating its first BITs with Egypt and Panama (Sachs 1984). After this, the United States negotiated eight additional BITs in the next four years (Vandevelde 1993, ). Those treaties were with Senegal, Haiti, the Democratic Republic of the Congo, Morocco, Turkey, Cameroon, Bangladesh, and Granada. After a hiatus, the program was given new life by the end of the Cold War (Vandevelde 1993, 630). As Eastern Bloc states moved towards opening their markets, the United States began negotiating a BIT with the Soviet Union, and with several formerly communist states, such as Poland. When the Soviet Union disintegrated, the United States quickly negotiated BITs with a number of former Soviet Republics. At the same time, the United States also continued to negotiate BITs with countries in Africa, Asia, and South America. As of December 2013, the United States had signed forty-seven BITs. Of the forty-seven treaties, forty-one had been ratified by both parties and gone into effect. A map of the countries the United States has signed a BIT with is shown in Figure 1. 7

10 B. Conventional Explanations for the United States BITs Program Although there have been a number of attempts to empirically examine why developing countries choose to enter into BITs (Elkins, Guzman, and Simmons 2006; Guzman 1998), there has been no attempt to empirically examine why the United States specifically or even developed countries generally decide to negotiate and sign BITs. Despite the lack of quantitative research on the topic, most commentators have claimed that the United States entered into BITs to influence the development of Customary International Law by reinforcing international legal principles on the treatment of foreign capital and to secure the rights of American investors (Elkins, Guzman, and Simmons 2006, ; Vandevelde 2005, 171; Lang 1998, 457; Vandevelde 1998a, 201-2; Vandevelde 1988,1-2; Gann 1985, 374). In other words, these scholars contend that the United States motivations were straightforward: investment treaties were negotiated to promote and protect investment. 4 This explanation has been put forth not only to explain the motivations of the United States, but also to explain the motivations of developed countries more generally (Tobin and Busch 2010, 2; Hamilton and Rochwerger 2005, 1; Vandevelde 2005, 171; Salacuse 1990, 661). Given the fact that commentators have consistently described BITs as treaties that developed states use for the purpose of promoting and protecting investment, it is unsurprising that scholarship evaluating the success of BITs has focused almost exclusively on whether the agreements have been successful methods of attracting capital (Salacuse and Sullivan 2005). Although there have been a few efforts to study the effects of BITs along other dimensions for example their effect on litigation (Simmons 2013) or the likelihood that their ratification leads to Preferential Trade Agreements (Tobin and Bush 2010) these studies have still conceived of BITs as treaties created for economic reasons. The result is that scholars studying the United 4 The one exception to this narrative that this author is aware of is Vandevelde (1993). Although Vandevelde (1993) argues that the motivations of the program have shifted over time, and that some BITs are political, Vandevelde does not consistently make this argument in later writing. For example, Vandevede more recently argued that the motivation for the developed country to conclude [BITs] was to obtain protection for its foreign investors, and that [i]n the United States in particular, the decision to negotiate BITs was very much motivated by a desire to create a network of treaties adopting the standard of prompt, adequate, and effective compensation for expropriation... (Vandevede 2005, 171). And Vendevede started his recent book on U.S. investment law by claiming that the principal purpose of the [U.S.] BITs [program] has been to protect U.S. investment in foreign countries (Vandevede 2009, 1). 8

11 States BIT program have viewed these agreements as a tool to promote investment that should be evaluated based on how well they accomplish that goal (Gann 1985, ). C. Limits to Economic Explanations for BIT Formation Although this conventional explanation for the growth of the United States BITs program has been widely repeated, there is a great deal of qualitative evidence to suggest that the program was not motivated by a desire to promote the development of international investment law or protect American capital invested abroad. First, there is evidence that the government officials who negotiated BITs did not expect a large increase in Foreign Direct Investment (FDI) as a consequence of the agreements (Vandevelde 1998b, 524). In the words of one commentator, as veteran U.S. BIT negotiators have repeatedly pointed out, U.S. negotiators have routinely alerted prospective BIT Partners not to expect that BITs would necessarily increase such flows from U.S. investors... (Alvarez 2010, 621 n.69). Of course, it is possible that the negotiators still hoped that BITs would increase investment flows, or that BITs would protect existing and future investments even if the investments were not flowing at a higher rate. The importance of this admission, however, is that if U.S. negotiators were skeptical that specific investment treaties would lead to new FDI from U.S. based investors, it is plausible that the officials pursuing the BITs were motivated by reasons other than simply increasing protections for American citizens and corporations. Second, if the United States were hoping simply to promote international law that is favorable to investment and provide protections to American investors, the United States should theoretically be willing to negotiate, sign, and ratify BITs with any state that was willing to agree to the U.S. model BIT. This at least would be a reasonable conclusion given the twin beliefs held by American negotiators that BITs will not impact existing investment flows (Alvarez 2010, 621) and that they require the United States to only make redundant concessions (Gann 1985, 374). As the United States failure to ratify the BITs it negotiated with Panama and Haiti illustrate, however, this is far from the case. Both Panama and Haiti were among the first countries to sign BITs with the United States, but before the Senate approved the BITs, the governments in both countries were overthrown by regimes that the United States did not support (Vandevelde 1993, ). As a consequence, the United States did not ratify the agreements 9

12 with these countries because doing so was deemed inconsistent with the foreign policy objectives of the United States. In other words, the United States cared less about the potential to develop favorable international law while protecting American investments in Panama and Haiti places experiencing serious unrest where such investor protections might be thought to be uniquely important than it did about the politics of having the treaty in place with a government that it did not support. Third, it appears that there is little pressure from American interest groups on the United States to ratify the BITs that it has signed. 5 Perhaps the best evidence of this fact is it has taken an average of 1,259 days for BITs that the United States has signed to be ratified and go into effect, but all of the BITs that the United States Senate has ratified have been approved by unanimous voice votes. Moreover, not only have business interests not opposed the BIT program, but organized labor has also not opposed the program (in part because there was no evidence that BITs would promote outward investment) (Vandevelde 2009, 26). This suggests that the long delays that have occurred before BITs are ratified are not a result of political opposition, but instead occur because the ratification of BITs is not a priority. This information also makes it plausible to infer that investors do not aggressively apply pressure to the Senate to approve BITs so that they can enjoy the increased protections that purportedly motivate the creation of the agreement in the first place. Fourth, BITs have not had an effect on investment flows between the United States and the countries that it has negotiated these treaties with (Peinhardt and Allee 2012). When looking at BITs overall, and not just the U.S. BITs program, the evidence that BITs directly increase the flow of FDI between the two countries that have negotiated them is mixed (Shaffer and Ginsburg 2012, 36-38; Yackee 2010, ). In the last several years, there have been a number of studies showing that BITs do not have any direct positive effect on FDI (Yackee 2010; Gallagher and Birch 2006), whereas competing studies using different methodology that focuses on total increases in investment and not just bilateral increases have found that BITs do in fact have a positive effect on overall FDI (Neumayer and Spess 2005; Salacuse and Sullivan 2005). One 5 Appendix B presents information on the Senate s consideration of all BITs the United States has signed. Although a detailed analysis of the Senate ratification of U.S. BITs has the potential to provide interesting insight into the agreements more broadly, this author is unaware of any scholarship that has closely explored the subject. A full treatment of the topic, however, is beyond the scope of this paper. 10

13 recent commentary on the state of the scholarship on the topic concluded that only studies examining bilateral investment flows between BIT parties find that BITs have little impact, whereas studies focusing on overall investment flows into BIT parties find that they have positive effects (Shaffer and Ginsburg 2012, 37). When looking solely at the U.S. BITs program, however, the best existing evidence that this author is aware of is much clearer: U.S. BITs have not had a statistically significant influence on investment patterns between the United States and its treaty partners (Peinhardt and Allee 2012). Fifth, there is evidence that BITs might not influence investment decisions. In one recent study, Yackee (2010) compiled evidence from a number of unique sources to argue that BITs do not impact the investment decisions of U.S. companies. Specifically, this study provided evidence from a survey of general counsels to United States-based multinational corporations and found that these individuals did not believe that the presence of a BIT impacted their company s investment decisions (Yackee 2010, ). Moreover, Yackee also compiled survey evidence from providers of political risk insurance that found that those insurers do not factor the presence of a BIT into their underwriting decisions (Yackee 2010, ). These alternative sources of evidence suggest that American investors were not eager to take advantage of the protections that were allegedly being negotiated for their benefit. D. A Political Theory of BIT Formation Taken together, this evidence suggests the United States may have had other motivations for negotiating BITs generally, and for picking which countries to negotiate them with specifically. My theory is that, counter to the conventional narrative, the United States BITs program was not used primarily as a means of influencing the development of international investment law and protect American investments abroad, but instead as a means of improving relationships with politically important developing countries. The foundation of this theory is the idea that there are four features of BITs make them a particularly useful foreign policy tool from the United States perspective. First, BITs are inexpensive. Unlike other tools that can be used to improve alliances such as foreign aid BITs do not require the United States to outlay funds. Second, BITs require the United States to only make redundant promises (or at least they were initially seen by U.S. policymakers that way (Gann 1985, 374)). That is to say, investors with capital in the United States are already 11

14 given access to U.S. courts, and the government believed that it was unlikely to expropriate foreign investment in any event. 6 Thus, the promises extracted from the United States were things the government had already pledged to do and thus created no new obligations. Third, BITs are easy to sell domestically. To both the United States Congress and the public, these treaties can be presented as a way to ensure that American investors are protected and given the same legal rights abroad that America extends to foreigners. Fourth, there is a standard model in place so that negotiating additional BITs requires relatively little effort. 7 Given these attractive features, the United States should be willing to sign a BIT with a government it wishes to support when the foreign government would like to have a BIT with the United States. The obvious question, of course, is why the developing state would want a BIT with the United States instead of an alternative low cost benifit like a state visit or statements of support when the BIT imposed at least some obligations on the developing state. My theory is that foreign governments were interested in signing BITs with the United States because doing so had the potential of producing domestic political benefits in two ways. First, even though foreign states were warned that having a BIT with the United States may not result in increased investment, there was still uncertainty about what the actual economic impact of the agreement would be. Foreign governments were thus able at least to argue that the agreement had the potential to produce economic benefits, and that they were trying to take steps to improve their countries economic conditions. Second, although BITs are a relatively low cost tool for the United States, there are still some negotiating costs associated with them for the U.S. government. As a result, the fact that the United States was willing to expend some effort 6 Investment Protections in U.S. Trade and Investment Agreements: Hearings Before the Committee on Ways and means of U.S. House of Representatives, 11th Cong. 65 (2009) [hereinafter 2009 Hearings] (testimony of Linda Menghetti, Emergency Committee For American Trade) (noting that protections for foreigners are somewhat redundant in that they have very strong protections already in the U.S. law and Constitution. And when they do challenge it, what they find is, again, the U.S. provisions form takings to due process and transparency issues all incorporated in that dispute resolution process.... ). 7 Bilateral Treaties Concerning the Encouragement and Reciprocal Protection of Investment: Hearing Before the Senate Committee on Foreign Relations, 104 th Cong. 4 (1995) (testimony of Daniel Tarullo, Assistant Secretary of State for Economic and Business Affairs) ( The BIT Program, I think we should take good note, is a relatively low cost... [because]... BITs are negotiated on the basis of a prototype document and only minor changes to that prototype language are generally accepted. As a result, the program requires only modest negotiating resources. ). 12

15 negotiating a BIT and passing it through the Senate provides a signal that the U.S. government was supportive of the foreign government. The combination of these features made many countries in the developing world interested in forming the treaty with the United States. Since BITs produce a benefit for the foreign government, it would be reasonable that the United States would enjoy an improved relationship with that country in the short term after the agreement was signed. This is because states are engaged in repeated interactions. After the United States gave a political benefit to the foreign state, it would be prudent for the new treaty partner to reciprocate so that the United States would provide it other forms of favorable treatment in the future. The benefits that the treaty partner should be expected to provide to the United States, however, would likely be relatively modest. This is because signing a BIT likely would not produce large political benefits for the foreign government, and as a result, the foreign government would be unlikely to pay a large price for receiving it. The theory that the U.S. BITs program was motivated by political considerations generates two clear testable hypothesizes. First, variables that measure the political importance of countries in the developing world should be better predictors of which countries the United States signed BIT with than variables that measure investment considerations. Second, signing a BIT with a developing country should produce modest political benefits for the United States in the short-term from the treaty partner. These hypotheses are tested in Parts III and IV. III. TESTING THEORIES OF BIT FORMATION To test the theory outlined in Part II, this section of the paper tests whether the set of countries that the United States has chosen to sign BITs with is better explained by investment considerations or political considerations. First, I explain the method used to test theories of BIT formation. Second, I discuss the data collected for this project. Third, I report the results of a series of statistical tests that suggest that investment considerations have not been determinants of which countries the United States has signed BITs with, but that whether a state was politically important was a determinate of BIT formation. Fourth, I present the results of a series of robustness checks that support my results. 13

16 A. Empirical Approach To empirically test my theory, I have constructed a time-series cross-sectional (TSCS) dataset for every potential BIT partner for the United States. Because the United States has exclusively signed BITs with developing states, I include an observation for every country that is not a member of the OECD for all years between 1981 the year when the United States started negotiating BITs and 2009 when the U.S. signed its most recent BIT with Rwanda. 8 The dependent variable takes the value 0 if the two countries have not signed a BIT in year t, and 1 if they have signed a BIT in year t. After a country has signed a BIT with the United States, it drops from the dataset. The data is thus structured for a survival analysis of the likelihood of the onset of a BIT. Researchers generally, and international relations scholars specifically, have struggled with how to model survival analyses correctly when using TSCS data (Beck et al. 1998). The problem is that logit and probit models assume temporal independence between observations while hazard models make an assumption that there is temporal dependence between units, and selecting an approach based on a false assumption will lead to an incorrect estimation of standard errors. To address this problem, I use the method suggested by Carter & Signorino (2010) and model BIT onset using logit regression, but control for the possibility that the observations were temporally dependent by including three variables Time, Time 2, and Time 3 that account for the time that elapsed since the beginning of the BITs program. 9 Additionally, to account for autocorrelation and heteroskedasticity, I calculate robust standard errors clustered by country. Using this approach to model BIT formation has three advantages. First, by using logit models, the results produced are more familiar and easier to interpret than hazard ratios, but still account for temporal dependence. Second, using logit models makes it possible to perform a number of robustness checks that are not available with hazard models a fact that will be exploited in Part III.D. Third, this approach is consistent with other recent empirical scholarship 8 The reason I exclude members of the OECD from the dataset is that OECD membership is an excellent proxy for the wealthy, developed, industrialized nations that were not potential BIT partners for the United States. The one country that was a member of the OECD and later signed a BIT with the United States is Turkey. As a result, I still included Turkey within the dataset. 9 Specifically, for the variable Time, observations in 1981 are coded as 1, 1982 as 2, etc, whereas Time 2 are the values squared and Time 3 are the values cubed. 14

17 in international law (Verdier and Voeten 2013) and comparative law (Ginsburg and Versteeg 2014). B. Data As previously noted, the dependent variable in this analysis is whether a BIT was signed between the United States and a potential treaty partner in a given year. Although by most counts the United States has signed 47 BITs, 46 BITs are analyzed in this study because the BITs with the Czech Republic and Slovakia were negotiated and signed with Czechoslovakia, but then simply inherited by each successor state after their split. As a result, Czechoslovakia drops from the dataset after signing a BIT with the United States, and does not reenter after the split. The independent variables collected for this study are all lagged one year and fall into three categories. 10 First, two independent variables are used to test whether BITs are a product of a desire to protect investment. The first of these variables is Investment Protection. This variable is from a proprietary dataset of investment risks developed by the Political Risk Services Group that has been widely used in academic research (perhaps most notably in Acemoglu et al. 2001). The variable measures countries annual investment risk on a twelve-point scale from 1 (lowest investment protection) to 12 (highest investment protection). If the U.S. BITs program was motivated by a desire to improve the climate for American investments, then Investment Protection should have a negative relationship with the likelihood of a BIT being signed. 11 The second variable is the natural log of the annual US FDI Outflows of foreign direct investment from the United States to the potential partner country each year. This variable was taken from the historical dollars of FDI data maintained by the U.S. Bureau of Economic Research, and was converted to constant dollars. 12 If the United States was motivated to improve protections for existing investments, then US FDI Outflows should have a positive effect on the likelihood of a BIT being signed. Second, two independent variables are used to test whether a counties political importance to the United States drove BIT formation. The first is the natural log of the amount of 10 Appendix C presents a correlation matrix of the independent variables. 11 It may be reasonable to hypothesize that there is a non-linear relationship between Investment Protection and BIT onset. This possibility is explored in Part III.D. 12 This data is available at < table of inflation adjustment factors used is available at < 15

18 Military Aid received from the United States in a given year. Previous research has suggested that military aid is a strong proxy for political importance to the United States because the recipients of military aid are perhaps the most crucial states to U.S. foreign policy objectives (Meernik, Krueger, and Poe 1998; Poe and Meernick 1995). If the U.S. BITs program was motivated by political considerations, Military Aid should have a positive effect on the likelihood of a BIT being signed. The second variable that measures a country s political importance is whether the country was a Former Communist state. 13 At the end of the Cold War, securing and improving relationships with countries from the former Eastern Bloc was a major foreign policy goal of the United States. In fact, in the Support for Eastern European Democracy (SEED) Act of 1989, Congress specifically urged the President to sign BITs with Poland and Hungary. 14 Although the United States likely also hoped to open up these markets for investment, improving relationships with these countries was a clear priority independent of any economic concerns. As a result, if the U.S. BITs program was motivated by economic factors, Former Communist should have a positive effect on the likelihood of a BIT being signed (even after conditioning on investment considerations). Third, a number of control variables were collected for this project to account for other considerations that may influence the United States decision to sign a BIT with another country. 15 First, because it has been hypothesized that whether a country is a democracy may influence that nation s willingness to ratify a BIT competing theories have been advanced that democracies may either be more willing to ratify BITs because they already have strong commitments to property rights, or less likely to ratify BITs because the promises are 13 This variable is coded as 0 in year t if a country is either communist or has never been communist, but is coded as 1 in year t if a state was formerly communist but has transitioned to be non-communist USC 5401(c)(15). 15 Over the last decade, there have been a number of influential studies that have attempted to explain BIT formation (Allee and Peinhardt 2010; Tobin and Busch 2009; Elkins, Guzman and Simmons 2006; Guzman 1998). Instead of trying to include all of the variables that have been used by these studies, I focus on controlling on a few widely used and theoretically relevant variables for two reasons. First, including every possible control variable would create a garbage can regression that could produce biased estimates (Achen 2005). Second, the methodological approach used in the second half of this paper results in a dataset of 92 observations, but requires conditioning on 5 years of values for each variable. Since this severely limits the number of covariates that can be used in the analysis, I have chosen to use a limited number of control variables in this part of the analysis to ensure consistency across the paper. 16

19 redundant I control each country s Polity Score as a measure of democratization (Marshall and Jaggers 2011). Polity Score is a widely used variable that places countries on a scale from -10 (most autocratic) to 10 (most democratic). Second, since a country s level of development may influence its attractiveness as a potential BIT partner, I control for the log of a country s GDP Per Capita. This variable was coded using the 2012 Penn World Tables (Heston et al. 2012). Third, since countries that are taking steps to liberalize their economy may be more desirable BIT partners, I control for whether a country has Open Capital Accounts as a proxy for financial openness. I specifically use the index developed by Chinn and Ito (2008) that codes countries relative openness and intensity of capital account restrictions using the IMF Annual Report on Exchange Arrangements and Exchange Restrictions. Fourth, since the United States may sign BITs in response to the potential partner s relationship with other countries, I control for the number of Prior BITs that the partner country previously signed. This variable was coded based on the ICSID database of BITs. Finally, it is important to note that, as is typical when using variables from different sources, there were many observations with missing values for the independent variables. Simply dropping these observations via listwise deletion, however, would bias the results of this study because the missing observations are likely non-random (Honaker and King 2010). To account for this source of bias, I imputed values for the missing observations using the Amelia II package for the R programing language (Honaker, King, and Blackwell 2011). Imputing values for missing observations has been shown to produce more reliable results (Honaker and King 2010), and is also consistent with the practice of recent international relations scholarship using timeseries cross-sectional data (Lupu 2013a; Lupu 2013b; Nielsen et al. 2012; Hill 2010). C. Results Figure 2 graphically presents the results of two logit regressions that test the influence of investment considerations Investment Protection and US FDI Flows and political considerations Military Aid and being a Former Communist on the likelihood that the United States will form a BIT with a country in a given year. 16 The base model includes only the four key variables of interest, whereas the other model includes the control variables discussed in the 16 The regressions that Figure 2 is based on are Models (1) and (2) in Appendix D. 17

20 previous section. The results present are the simulated first differences as the key explanatory variables moves from their minimum to their maximum value, while all other variables are held at their means. 17 Figure 2 thus reports the influence that the variable of interest has on the percent change in likelihood that a BIT will be formed in a given year. Figure 2: Marginal Effects on Likelihood of U.S. BIT Formation Investment Protection Base Model With Controls US FDI Flows Base Model With Controls Military Aid Base Model With Controls Former Communist Base Model With Controls Marginal Effects on BIT Signing (%) Figure 2 depicts the estimated impact of key explanatory variables on the likelihood that the United States will sign a Bilateral Investment Treaty with a foreign state in a given year. The results presented are the first differences as variables change from the minimum to maximum value, with all variables held at their means. Point estimates are represented by dots, with the lines representing 95% confidence intervals. Statistically significant estimates are shown with solid lines, while statistically insignificant estimates are shown with dotted lines. 17 For a defense of presenting regression results graphically using this method, see King, Tomz, & Wittenburg (2000) and Kastellec & Leoni (2007). 18

21 As the results in Figure 2 clearly show, the two independent variables that measure investment considerations do not achieve statistical significance. Investment Protection does have the hypothesized negative effect but is not close to achieving statistical significance. The second variable that measures investment considerations US FDI Outflows also does not have a statistically significant influence on the likelihood of BIT formation. In both the base model and the model with control variables, the effect of US FDI Outflows is negative, but the effect is both substantively small and far from conventional levels of significance. These finding suggests that United States has not selected BIT partners based simply on whether the country presented a particularly risky investment climate for American capital exporters, or the amount of capital that the United States is currently sending to the country. In contrast, the two variables that measure political considerations Military Aid received from the United States and being a Former Communist state both have a positive and statistically significant effect on the likelihood of BIT formation. Military Aid is statistically significant at the 0.05 level for the base model, and the 0.10 level for the model that includes control variables. Additionally, Former Communist states is statistically significant at the level for both the base model and the model including control variables, and also has the largest substantive effect of the four independent variables explored in this paper. Together, these results lend strong support to the theory that the United States was driven to sign BITs with countries because of political considerations. It is also worth considering the size of these effects. Military Aid increases the chance that the United States would form a BIT with a partner country in a given year by 1.0 percent in the base model, and 1.1 percent in the model with control variables. This effect is even larger for Former Communist, which increases the probability by 7.1 percent in the base model, and 6.0 percent in the model with control variables. Although the size of these effects is small, the United States signed less than 50 BITs in 30 years, so the likelihood that a BIT would be signed with a country in any given year is incredibly low. To put the magnitude of these effects in perspective, it is possible to calculate a baseline probability that a country will sign a BIT with the United States each year. To do so, I estimated a model that includes the two variables measuring investment considerations and the two variables measuring political considerations, as well as all four control variables, and then estimated the probability of a BIT being formed with 19

22 all variables set at their means. 18 The baseline probability that the United States would sign a BIT with a country in a given year is just 0.5 percent. As a result, increasing the probability of a BIT signing by between roughly 1.0 and 4.0 percent, as the political variables do, is a relatively large effect given the low baseline probability of BIT formation. These results thus suggest that there is only weak empirical support for the idea that the United States signs BITs with developing countries because of investment considerations, but that the political importance of a country does have explanatory power. Although these results do not provide definitive proof that BITs are motivated by foreign policy considerations, they at least suggest that foreign policy considerations are better predictors of BIT formation than investment considerations thus lending clear support to the theory outlined in Part II. D. Robustness Checks The results presented in the previous section are also robust to a series of alternative model specifications. Specifically, the results are robust to estimating both the base model and the model with control variables using a variety of alternative specifications. 19 First, one potential concern with the analysis in the previous section is that there are only a small number of BITs signed relative to the number of country-years analyzed. Since the analysis uses an onset structure, countries are dropped from the dataset in the year after a BIT is signed. The result is that there are 46 observations that take the value of 1 and nearly 3,700 observations that take the value of 0. When the onset of an event is rare, previous research has shown that it can result in biased estimates of the probability of that event occurring (King and Zeng 2001). To ensure that this problem did not bias my results, I estimated a Rare Events Logit model (King and Zeng 2001). The results produced using this method are consistent with the results presented in the previous section. The coefficients for the variables of interests are significant at the same levels, and have substantively similar effects. It is thus reasonable to conclude that the results in the previous section were not biased because the onset of a BIT is rare. Second, a major concern with any TSCS analysis is that there is serial correlation and underlying heterogeneity between the groups being analyzed in my case, countries. Although I use the standard approach of correcting for this possibility by calculating robust standard errors 18 This model used to estimate this baseline effect is Model (2) in Appendix D. 19 The results of these robustness checks are reported in Appendix D. 20

23 clustered on country, recent research has suggested that this approach may not adequately account for heterogeneity between groups in the data (King and Roberts 2014). As a result, I have estimated a logit model with country random effects. This approach has the advantage of directly modeling the underlying heterogeneity, instead of simply correcting the standard errors after the fact (Nielsen 2013). When using this method, the variables that measure investment consideration still do not achieve statistical significance, but the variables that measure political considerations still have a statistically significant positive effect on BIT onset. Third, it may be reasonable to hypothesize that Investment Protection has a relationship with BIT onset, but that it is a non-linear relationship. Specifically, it may be the case the United States would not be interested in signing a BIT with countries that have the lowest levels of investor protection, but also would not feel the need to sign a BIT with countries that have the highest levels of investor protection. To test this possibility, I estimated my initial models while also including the quadratic term for Investment Protection (Investment Protection 2 ). Doing so did not reveal a statistically significant relationship between either Investment Protection or Investment Protection 2, but the effect of Military Aid and Former Communist remained statistically significant when including Investment Protection 2 in the regressions. Fourth, it may be possible to argue that treating the variable Former Communist as a measure of a country s political importance is inappropriate for two different reasons. First, it may be reasonable to think that the BITs negotiated with the Eastern Bloc may have had unique motivations, and thus it would be inappropriate to conclude that investment considerations cannot explain the BIT program overall when these countries where include in the data. To test this theory, I estimated the base model and model with controls after dropping states that were formerly communist from the dataset. In both models, neither variable measuring investment considerations achieves statistical significance, but Military Aid still has a positive effect that is statistically significant at the 0.05 level. This result suggests that, even excluding the wave of BITs negotiated with the Eastern Bloc in the early nineties, political considerations are a better predictor of BIT onset than investment considerations. Second, it may be reasonable to argue that investment considerations influenced with which Former Communist states the United States was most likely to sign a BIT. To test this possibility, I estimated a series of regressions in which I interacted a different explanatory variable with the Former Communist variable. In these regressions, the interaction between Investment Protection and Former Communist as well as 21

24 the interaction between US FDI Outflows and Former Communist is not statistically significant. There is, however, a negative relationship for the interaction between Military Aid and Former Communist that is statistically significant in the model with control variables. This is likely because the United States was eager to form BITs with communist states quickly after their transition and thus in many cases before they were large recipients of military aid. Fifth, it could be the case that investment considerations influence which countries the United States signs BITs with, but that the country must first be an ally before even being considered for a BIT. To test this theory, I estimated the base model and model with controls after dropping observations where the state that did not receive any military aid from the United States in the prior year. In both models, neither variable measuring investment considerations achieves statistical significance, but Former Communist still has a positive effect that is statistically significant at the level. Sixth, until now this paper has analyzed the onset of BITs being signed as the dependent variable. There are five BITs, however, that the United States has signed that have never gone into effect. 20 It might thus be the case that different factors determined the onset of BITs actually going into effect. To test this, I estimated regressions using the onset of BITs going into effect between the United States and partner countries as the dependent variable. The results of doing so are consistent with the initial results. Taken together, the results of these robustness checks lend strong support to the theory that political considerations influenced which countries the United States was willing to sign a BITs. The Military Aid and Former Communist variables both have a consistently statistically significant and positive effect on the likelihood that the United States would sign a BIT with a given country. In contrast, there is little evidence that investment considerations had any effect on the BITs that the United States signed: Investment Protection and US FDI Flows failed to achieve statistical significance in any of the models estimated. The results thus suggest there is much better evidence that the United States was motivated to sign BITs with countries based on their political importance, and not their investment profile. 20 These BITs are with Haiti, Russia, Belarus, Uzbekistan, Nicaragua, and El Salvador. 22

25 IV. TESTING THE POLITICAL CONSEQUENCES OF BITS The second step to testing my theory is to examine whether BITs have produced political dividends for the United States. This part of the paper does exactly that. First, I describe the empirical approach that was used for this analysis life history matching to try to produce credible causal estimates of the political impact of BITs. Second, I discuss the three dependent variables that measure political support for the United States that are used for this analysis. Third, I report the primary results of this analysis, and fourth, I discuss the results of a series of robustness checks that support these findings. A. Methodological Approach The difficulty in measuring the political benefits of BITs is that it requires finding a quantitative method that makes it possible to estimate the causal effects of BITs on state-to-state relationships while controlling for selection effects. Roughly fifteen years ago, scholars began to empirically study the effect of treaty ratification on state behavior (Simmons 1999; Hathaway 2002). Although these early efforts were an important step forward that improved our understanding of international legal agreements, many of these early efforts also had flaws in their research design (Goodman and Jinks 2003; von Stein 2005). One specific problem was that these efforts were not adequately able to control for selection effects (Hill 2010, ). That is to say, countries that had chosen to ratify treaties were systematically different than countries that had not, making it impossible to attribute observed differences in behavior to the treaties. In an effort to help correct for these problems, a second wave of empirical scholarship on international law has employed a variety of increasingly sophisticated quantitative methods (von Stein 2005; Simmons and Hopkins 2005; Simmons 2009; Hill 2010; Lupu 2013a; Lupu 2013b). One method that has been particularly popular is the use of matching. Although there are a variety of matching methods available (King et al. 2012; Ho et al. 2007), the basic intuition is that, if two observations are paired together that are as similar as possible in every way except that one has received a particular treatment for example, ratified a treaty then systematic differences in a dependent variable of interest can be attributed to that treatment. Although matching has allowed scholars to make considerable advances in understanding the influence of treaties on state behavior, matching time-series cross sectional data using 23

26 conventional methods poses two problems that researchers have almost entirely chosen to ignore. First, conventional matching methods do not account for trends in data (Nielsen, n.d.). The problem is that two countries might have the same value for a variable (for example, their level of democracy) in a given year, but the countries could be trending in opposite directions (that is, one country could be in the process of becoming more democratic while the other is sliding towards autocracy). Conventional matching methods simply pair countries with similar values of a variable in a given year, and thus are unable to take these trends into account. Second, conventional matching methods introduce post-treatment bias. The problem is that, when trying to estimate the causal effect of a given treatment, it is inappropriate to condition on covariates that are measured after the treatment, because doing so may block the causal pathways that allow the treatment to influence the outcome of interest (Blackwell and Glynn, n.d.). Conventional matching methods introduce post-treatment bias because they find matches for treated observations using values for covariates measured the year of each observation, even if that observation occurs several years after the treatment. To overcome these problems, I use a recently developed method that offers a promising way to estimate the causal effects of international agreements that both accounts for trends in data and also avoids introducing post-treatment bias. In a forthcoming paper, Nielsen and Simmons (2014) introduce the idea of using treaty ratification episodes as the unit of analysis. They define these episodes as the 11-year period surrounding ratification. For example, if Poland ratified a particular treaty in 1990, the ratification episode would be from 1985 to Nielsen and Simmons then match each ratification episode to an equally long non-ratification episode using a new technique life history matching that makes it possible to match on multiple years of data instead of one (Nielsen, n.d.). 21 Nielsen and Simmons (2014) specifically match observations using the first five years of data from each episode, and examine the effect of the 21 Life history matching is a technique first developed in Nielsen (n.d.). The basic insight is that for each observation that appears in a dataset, that the values of covariates for multiple prior years are collected. For example, if a covariate of interest in a TSCS dataset is GDP Per Capita, separate variables would be created that are: GDP Per Capita T, GDP Per Capita T-1, GDP Per Capita T-2,, etc. Then if an observation in the dataset is Brazil 1999, Brazil s GDP Per capita for 1999 would be stored in GDP Per Capita T, Brazil s GDP Per capita for 1998 would be stored in GDP Per Capita T-2, Brazil s GDP Per capita for 1997 would be stored in GDP Per Capita T-3, etc. Each of these variables is then included as a separate term during standard matching procedures. 24

27 dependent variable of interest in the six years during and after ratification. In the Poland example, Poland may be matched with Hungary based on their values for covariates measured in 1985 through 1989, and then differences in the dependent variable of interest from 1990 to 1995 would be analyzed. Since observations are matched on five years of control variables that all occur prior to a treaty being formed, this approach is able to account for trends in the data without introducing post-treatment bias. Given the advantages of this approach, I follow Simmons and Nielsen (2014) and use BIT Formation Episodes as my unit of analysis, and match these episodes using life-history matching. B. Data In addition to the data discussed in Part III.B., three additional dependent variables were collected for this study. One difficulty of this project is finding dependent variables that measure the strength of relationships between two states in a way that makes it possible to detect a benefit for the developed state. That problem is that material rewards of treaty ratification are likely to flow from developed to developing states. For example, Rwanda may expect that signing a BIT with the United States will lead to being viewed more favorably when it is time for America to decide how to distribute foreign aid dollars (Salacuse 2010, 442 n.75), but it would be patently unreasonable for the United States to expect aid from Rwanda. That said, this does not mean that the United States does not receive a political benefit from negotiating a BIT with Rwanda, it simply means that the political benefits may be difficult to measure. The challenge is thus to find dependent variables that capture the strength of a geopolitical relationship where the developed state would be likely to be on the receiving end of the benefit. Three dependent variables that fit this criterion have been identified for this study. The first is a measure of the similarity in United Nations General Assembly voting between the United States and the BIT partner. Voting in the UN is a public action that, although it may often be purely symbolic, is at least possible to clearly measure as a proxy of closeness in preferences and policies between states. Given these properties, previous international relations scholarship has used UN voting as a way to directly measure changes in the relationships and alliances between states over time (Voeten 2004; Voeten 2000; Gartzke 2000; Gartzke 1998). One specific variable that has recently been developed to measure similarities between countries 25

28 in their UN voting is the difference in their UN Ideal Points (Baily, Strezhnev, and Voeten 2013). This variable is calculated by first using spatial modeling techniques to estimate each country s ideal point on votes taken in the General Assembly in a given year, and then estimating the difference in ideal points between two countries in that year. The more similar two countries voting patterns are, the smaller the difference in their ideal points will be. Although there are certainly problems with using UN Ideal Points to measure the strength of two countries relationship, 22 the advantages of the measure still outweigh the drawbacks (Gartzke 1998, 14). Specifically, all UN Members vote in the General Assembly in every year, which makes it possible to look at changes in patterns over time. Moreover, UN voting occurs on a range of topics in each year, meaning that UN voting is a broad-based measure of the similarities in countries preferences. Finally, it has even previously been hypothesized that BITs may be negotiated in part as a signal about future intentions in UN voting (Alvarez 2010, 621). As a result, UN Ideal Points provides a promising way to measure year-to-year changes in the United States strength of relationships with treaty partners that are a consequence of signing BITs. The specific dependent variable I use is the six-year average of the difference in UN Ideal Points between the United States and other countries. The second dependent variable used to test whether the U.S. BITs program has produced political dividends is the number of United States troops deployed in a country in a given year. There has been a line of research examining the effects that the United States commercial relationships with developing countries have had on the likelihood that the country will be willing to allow U.S. troops to be deployed on their soil (Biglaiser and De Rousen 2009; Biglaiser and De Rousen 2007). As Biglaiser and Rousen (2009) demonstrated, the United States often must provide developing countries with economic incentives to be able to later station troops within their borders. These incentives help to soften domestic opposition to the presence of U.S. troops. Trade concessions are thus a strong predictor that U.S. troops will later be 22 It would be reasonable to argue that UN Voting is a somewhat crude and noisy measure of the strength of relationships for at least four reasons. First, there are many determinants of a country s UN voting. Second, countries decide how to vote in the UN by evaluating far more factors than simply whether a country with which they share a strong relationship would prefer for them to vote one way or the other. Third, UN voting might be a symbolic action that does not reflect other foreign policy preferences that are more important to a country. Fourth, UN voting may be driven in part by the issues presented in that year, making trends a function of the subject matter and not a function of changes in underlying preferences. 26

29 stationed within a given country. It would thus be reasonable to hypothesize that signing a BIT with a developing country would make it more likely that that state will later allow U.S. troops on its soil. Using data from Kane (2005), the dependent variable used to test this is the six-year average number of troops the United States has stationed within a given country. The third dependent variable used to test whether the U.S. BITs program has produced political dividends is whether the partner country was a member of the Iraq War Coalition in At the time of the invasion of Iraq in March 2003, many traditional allies of the United States in the developed world such as Canada, France, and Germany were not members of the coalition supporting the war. Instead, the countries comprising the coalition were frequently developing states. Prior research has suggested that these developing states were members of the coalition because of their economic and strategic linkages with the United States (Newnham 2008). As a consequence, one test of whether BITs have been successful at improving relationships with countries in the developing world is whether the prior presence of a BIT resulted in a country being more likely to be part of the Iraq War Coalition. This dependent variable is a dummy variable for the year 2003 only, and was based on information released by the White House on March 27, C. Results Following the procedure discussed in Part IV.A, the first step in my analysis was preprocessing the data using life history matching. The specific matching procedure I used was nearest neighbor propensity score matching, implemented using the Matchit package for the R programing language (Ho et al. 2009). 24 BIT Formation Episodes were matched with nonformation episodes based on five years worth of values for all of the independent variables used in Part III: Investment Protection, US FDI Outflows, Military Aid, Former Communist, Polity 23 Appendix E provides a complete list of the members of the Iraq War Coalition. 24 Following Nielsen and Simmons (2014), the matching produce retains all of the treatment episodes. The treatment episodes are each matched to one control unit, and the matching is greedy (which means that only control episode can only be used once). I also follow Nielsen and Simmons (2014) by allowing for overlapping non-ratification to come from the earlier history of the same country, but not allowing for overlapping ratification and nonratification episodes. Instead, ratification episodes take precedence. Since this allows for multiple observations to be from the same country, I account for any non-independence between the observations by calculating robust standard errors clustered by country. 27

30 Score, GDP Per Capita, Open Capital Acts, and Number of Prior BITs. A matched dataset was created for each dependent variable, and the balance statistics from the matching procedures are reported in Appendix F. 25 The use of matching reduced the covariate imbalance between the two groups and thus makes it possible to produce less-biased estimates of the treatment effect of having a BIT with the United States. After matching, the next step was to run a series of linear regressions to estimate the effect of BITs on the three dependent variables that measure the potential benefits the United States may have received from the treaties. For each of the three dependent variables, a regression was estimated using the matched dataset. To account for the possibility of any remaining imbalance, I followed the standard procedure of including the treatment variable (having signed a BIT with the United States) as well as all of the control variables that the observations were matched on in the post-matching regressions (Ho et al. 2007). The results of this analysis are reported in Figure 3. Figure 3 graphically presents the marginal effects of these regressions for the treatment variable having signed a BIT with the United States with all other variables held at their means. Given the small sample sizes, Figure 3 presents 90% confidence intervals The matched datasets produced for UN Ideal Points and Troop Deployment are identical, but the matched dataset for the Iraq War Coalition is different because it excludes observations where the focal year was after Appendix G presents the results reported in Figure 3 in a standard regression table. 28

31 Figure 3: Estimates of the Political Impact of BITs Signed by the United States UN Ideal Points Troop Deployment Iraq Coalition (p = 0.013) -1, ,500 (p = 0.762) (p = 0.055) Figure 3 depicts the estimated impact of the United States signing a Bilateral Investment Treaty with a foreign state on three political outcomes. Each estimate is from a separate regression. The results presented are the first differences as variables change from the minimum to maximum value, with all other variables held at their means. Point estimates are represented by dots, with the lines representing 90% confidence intervals. Statistically significant estimates are shown with solid lines, while statistically insignificant estimates are shown with dotted lines. As Figure 3 shows, having signed a BIT with the United States has a negative effect on the difference in UN Ideal Points between the United States and the partner country that is statistically significant at the 0.05 level. More specifically, having a BIT with the United States makes a country vote an average of 0.33 points lower than the similar countries that comprise the control group during the six years during and after a BIT is signed with the United States. Obviously, it is important to put the magnitude of this effect in context. This variable is measured on a scale of 0 to 5, with an average value of 3.10 and a standard deviation of 0.71 across the entire dataset used for this project. Or, perhaps more helpfully, between 1981 and 1991, Russia had an average difference in UN Ideal Point with the United States of 4.1, and an average difference of 2.1 between 1992 and In other words, the end of the Cold War resulted in a difference of roughly 2.0 and the estimated effect of a BIT is That said, it is worth noting that, although the size of this effect is modest, it is statistically significant at the 0.05 level despite the fact that the treatment and control groups are intentionally designed to be as similar as possible along a number of observable relevant dimensions, with the exception that the treatment group has signed a BIT with the United States. Given that the most recent 29

32 empirical study of the investment consequences of the United States BITs program (that this author is aware of, at least) did not find even a small change in investment patterns resulting from U.S. BITs (Peinhardt and Allee 2012), this is perhaps a surprising result. Second, Figure 3 shows that having signed a BIT with the United States has a negative effect on the average number of U.S. Troops deployed in the partner country. Countries that have signed a BIT with the United States have an average of 174 troops less a year during the six year period during and after ratification. This result, however, is far from being statistically significant. Since the result is statistically indistinguishable from zero, little attention should be paid to either the direction or magnitude of this effect. Third, Figure 3 shows that having signed a BIT with the United States has a positive effect on the likelihood that the partner country would participate in the Iraq War Coalition. In fact, countries that have signed a BIT with the United States were 18% more likely to participate in the Iraq War Coalition, and the result is statistically significant at the 0.1 level. Although this result constitutes evidence of the BITs producing a major political benefit for the United States, the result is not robust to a single one of the alternative model specifications discussed in the next section, and as a result, it would likely be a mistake to place much faith in its validity. D. Robustness Checks To test the robustness of these results, I also estimated a series of regressions without using the matching procedure employed in the previous section. The results of this analysis are reported in Figure Just like Figure 3, Figure 4 graphically presents the marginal effects of having signed a BIT with the United States, with all other variables held at their means. 27 Appendix G presents the results reported in Figure 4 in a standard regression table. 30

33 Figure 4: Robustness Checks of the Political Impact of BITs Signed by the United States 0 (A) Controls - 5 Years UN Ideal Points Troop Deployment -1, ,500 0 Iraq Coalition (p = 0.001) (p = 0.638) (p = 0.280) 0 (B) Controls - 1 Year , , (p = 0.001) (p = 0.612) (p = 0.294) 0 (C) Fixed Effects , ,500 (p = 0.005) (p = 0.423) Figure 4 depicts robustness checks estimated impact of the United States signing a Bilateral Investment Treaty with a foreign state on three political outcomes. Each estimate is from a separate regression. The results presented are the first differences as variables change from the minimum to maximum value, with all other variables held at their means. Point estimates are represented by dots, with the lines representing 90% confidence intervals. Statistically significant estimates are shown with solid lines, while statistically insignificant estimates are shown with dotted lines. First, even if using matching to pre-process the data improves the balance between the treatment and control group, one reasonable concern to have is that the matching procedure used may bias the results because of the particular observations that were discarded. As a result, I recreated my BIT Formation Episode research design, only did not pre-process the data by matching. To do so, I estimated a series of regressions that included one observation for each 31

34 ratification episode, but also included all possible non-ratification episodes. The results of this analysis are reported in Panel A of Figure 4. These results suggest that having signed a BIT with the United States has a statistically significant negative effect on the difference between the United States and the partner country s UN Ideal Point that is nearly identical in magnitude to the estimate produced via matching, but that is statistically significant at the 0.01 level. When using this approach, however, having signed a BIT with the United States did not have a statistically significant effect on either Troop Deployment or participation in the Iraq Coalition. Second, another possible concern with the empirical approach that I have used is that choosing to condition on five years worth of data for all of the control variables may have biased the results. To account for this possibility, I instead use a seven-year BIT Formation Episode, where I condition on one year of data, and then analyze the impact of BITs in the six years during and after BIT formation. As with the previous robustness check, I include one observation for each BIT Formation Episode, and all possible non-bit Formation Episodes. The results of this analysis are reported in Panel B of Figure 4. These results are nearly identical to the results reported in Panel A: having signed a BIT with the United States has roughly the same statistically significant, negative effect on the difference in UN Ideal Points, but does not have a statistically significant effect on either of the other two dependent variables. Third, another concern with the approach that I have used is that my estimates may be biased due to failure to account for variables that influenced both the treatment and dependent variable. Since it is certainly possible that both unobserved and observed factors are influencing my results (after all, perhaps the greatest flaw of using matching to account for selection effects is that it is possible to match observations only on observables), I also employ a standard regression model that estimated the impact of having signed a BIT with the United States while controlling for country and year fixed effects. 28 This model employs a standard TSCS design, where every country year in the dataset is included, and instead of the dependent variable being a six-year average, the dependent variable is simply measured in the year of each observation. This approach has the advantage of ensuring that there are not observed features of the countries that signed the BITs or years in which they were signed that are driving the results. The results of this 28 The downside of this approach is that it does not allow controlling for variables that vary over time like Polity Score or GDP Per Capita because their inclusion would introduce post-treatment bias. As a result, the regressions reported in Panel C include country and year fixed effects, but do not include any control variables. 32

35 analysis are reported in Panel C of Figure Once again, having signed a BIT with the United States has a negative and statistically significant effect on the difference in UN Ideal Points. The magnitude of this effect is actually slightly larger than the size of the effect estimated using other methods. Also, as with the other models estimated, the effect of the treatment on U.S. Troop Deployment on the foreign country s soil remains statistically insignificant. Fourth, these results were all estimated by examining the political benefits the United States has received from signing a BIT with a partner country. It is also worth considering whether these results would be the same if the treatment variable was a BIT actually going into effect. I thus re-estimated the results produced in Figure 3 and Figure 4, but the treatment variable is a BIT between the United States and a partner country going into effect. 30 The results of this analysis reveal that there is not a statistically significant relationship between a BIT going into effect and any of the three dependent variables (the one exception is the UN Ideal Point is statistically significant at the 0.1 level in a single model). This suggests that any political closeness the United States may receive from a BIT as measured by the countries difference in UN Ideal Points occurs when the BIT is signed, but dissipates over time. Because the political benefits occur after the treaty is signed, but not after it goes into effect, these results lend support to my theory that the United States is motivated to sign BITs because doing so generates political benefits independent of any increase in investment or investor protection. In light of these robustness checks, the results in this part of the paper suggest that having signed a BIT with the United States was associated with countries voting more similarly to the United States in the United Nations, but likely did not make those countries more likely to have U.S. Troops on their soil or to have been a member of the Iraq War Coalition. It may then fair to say that the U.S. BITs program has produced modest political benefits, but (perhaps unsurprisingly) has not radically altered the national security policy of partner countries. 29 Since the Iraq War Coalition dependent variable is measured in only 2003, the structure of the data is not time-series cross-sectional. It is thus impossible to estimate a model for this dependent variable while including country or year fixed effects, and as a consequence, Panel C only includes results for the other two dependent variables. 30 Appendix H presents these results in a standard regression table. 33

36 V. CONCLUSION Scholars studying the United States BITs program have consistently argued that it was motivated by a desire to help promote the development of international law that is friendly to investment, and to help protect American investments abroad. This paper challenges that view of the U.S. BITs program. Instead, I argue that BITs have been a foreign policy tool that the United States has used when it wanted a low cost way to signal that it would like to improve its relationship with a specific developing country. As a result, political considerations should help to explain the countries with which the United States signed BITs; and if the BIT program has been effective, BITs should have produced political benefits. This paper rigorously tested that theory by both showing that political considerations are better than investment considerations at predicting BIT formation, and that BITs are associated with at least modest political benefits. Of course, a few caveats are in order. First, although the results presented in this paper suggest that investment considerations did not have a statistically significant impact on the likelihood of BIT onset, this does not mean that investment considerations did not play a role in specific cases. Instead, the results simply suggest that political considerations are better predictors of BIT onset on average. Second, the same is true of the change in UN Ideal Points after a BIT is signed. The result of this paper should not be taken to mean that a change occurred in every case, or that a change would occur with any country the United States signed a BIT with in the future. Third, these results should not be interpreted strictly causally. Although the results in this paper results show that having a BIT with the United States in effect is associated with a statistically significant decrease in the difference between the United States and the treaty partner s UN Ideal Point, this might not be caused by the formation of the BIT. The point of using matching and linear regression is to try to account for other factors that may have caused the change in the dependent variable, but it is always possible that there are other variables that were not controlled for in the model and which may not even be observable that are in part driving the results. Demonstrating that signing a BIT is associated with a change in political outcomes, even if it is not causal, does lend support to the overall argument advanced by this paper that the United States used BITs with the goal of improving relationships. With that said, these findings still show that commentators should alter the way they describe, discuss, and evaluate the United States BITs program. Not only do the results of this 34

37 paper suggest that the narrative that the United States signed BITs to protect and promote investment be modified, but they also suggest that looking solely at the effect of BITs on FDI flows and investor protections may be the wrong way to measure their effectiveness. Instead, this paper demonstrates that it is time to acknowledge the political motivations of the program, and to start evaluating the program at least partially on political dividends that may it have produced. 35

38 REFERENCES Acemoglu, Daron, Simon Johnson, and James Robinson The Colonial Origins of Comparative Development: An Empirical Investigation. American Economic Review 91: Achen, Christopher H Let s Put Garbage-Can Regressions and Garbage-Can Probits Where They Belong. Conflict Management and Peace Sciences 22: Akhtar, Shhayerah Ilias and Martin A. Weiss U.S. International Investment Agreements: Issues for Congress. Congressional Research Service R Allee, Todd and Clint Peinhardt Delegating Differences: Bilateral Investment Treaties and Bargaining Over Dispute Resolution Provisions. International Studies Quarterly 54:1-26. Alvarez, José E The Evolving BIT. Transnational Dispute Management 6. Alvarez, José E The Once and Future Foreign Investment Regime. In Looking to the Future: Essays on International Law in Honor of W. Michael Reisman, edited by M.H. Arsanjani et al, 607. Boston, M.A.: Martinus Nijhoff Publishers. Baily, Michael, Anton Strezhnev, and Erik Voeten Estimating Dynamic State Preferences from United Nations Voting Data. Working Paper. Beck, Nathaniel, Jonathan N. Katz, and Richard Tucker Taking Time Seriously: Time- Series Cross Sectional Analysis with a Binary Dependent Variable. American Journal of Political Science 42: Biglaiser, Glen and Karl De Rousen Following the Flag: Troop Deployment and U.S. Foreign Direct Investment. International Studies Quarterly 51(4): Biglaiser, Glen and Karl De Rousen The Interdependence of U.S. Troop Deployment and Trade in the Developing World. Foreign Policy Analysis 5(3): Carter, David B. and Curtis S. Signorino Back to the Future: Modeling Time Dependence in Binary Data. Political Analysis 18: Chinn, Menzie and Hiro Ito A New Measure of Financial Openness. Journal of Comparative Policy Analysis. 10: Coyle, John F The Treaty of Friendship, Commerce and Navigation in the Modern Era. Columbia Journal of Transnational Law 51: Elkins, Zachary, Andrew Guzman, and Beth A. Simmons Competing for Capital: The Diffusion of Bilateral Investment Treaties, International Organization 60:

39 Gallagher, K. P. and M. B. I. Birch Do Investment Agreements Attract Investment? Evidence from Latin America. Journal of World Investment and Trade 7(6): Gann, Pamela B The U.S. Bilateral Investment Treaty Program. Stanford Journal of International Law 21: Gartzke, Erik Kant We All Just Get Along? Opportunity, Willingness, and the Origins of the Democratic Peace. American Journal of Political Science 42(1):1-27. Gartzke, Erik Preferences and the Democratic Peace. International Studies Quarterly 44(2): Ginsburg, Tom and Mila Versteeg Why Do Countries Adopt Constitutional Review? Journal of Law, Economics, & Organization (forthcoming). Goodman, Ryan and Derek Jinks Measuring the Effect of Human Rights Treaties. European Journal of International Law 15(1): Guzman, Andrew T Why LDCs Sign Treaties That Hurt Them: Explaining the Popularity of Bilateral Investment Treaties. Virginia Journal of International Law 38: Hamilton, Calvin A. and Paula I. Rochwerger Trade and Investment: Foreign Direct Investment Through Bilateral and Multilateral Treaties. New York International Law Review 18(1):1-59. Hathaway, Oona A Do Human Rights Treaties Make a Difference? Yale Law Journal 111(8): Hill, Daniel W Estimating the Effects of Human Rights Treaties on State Behavior. Journal of Politics 72(4): Ho, Daniel E., Kosuke Imai, Gary King, and Elizabeth A. Stuart Matching as Nonparametric Preprocessing for Improving Parametric Causal Inference. Political Analyses 15(3): Ho, Daniel E., Kosuke Imai, Gary King, and Elizabeth A. Stuart MatchIt: Nonparametric Preprocessing for Parametric Causal Inference. Journal of Statistical Software 42(8):1-28. Kastellec, Jonathan P. and Eduardo L. Leoni Using Graphs Instead of Tables in Political Science. Perspectives on Politics 5: King, Gary and Margaret E. Roberts How Robust Standard Errors Expose Methodological Problems They Do Not Fix, and What to Do About It. Political Analysis (forthcoming). 37

40 King, Gary, Rich Nielsen, Carter Coberly, James Pope, and Aaron Wells Comparative Effectiveness of Matching Methods for Causal Inference. Working Paper. King, Gary, and Langche Zeng Logistic Regression in Rare Events Data. Political Analysis 9(2): King, Gary, Michael Tomz, and Jason Wittenburg Making the Most of Statistical Analyses: Improving Interpretation and Presentation. American Journal of Political Science 44: Lang, Jeffery Symposium on the International Regulation of Foreign Direct Investment: Keynote Address. Cornell International Law Journal 31(3): Lupu, Yonatan The Informative Power of Treaty Commitment: Using the Spatial Model to Address Selection Effects. American Journal of Political Science (forthcoming). Marshall, Monty G. and Keith Jaggers Polity IV Project: Political Regime Characteristic and Transitions, Meernik, James, Eric L. Krueger, and Steve C. Poe Testing Models of U.S. Foreign Policy: Foreign Aid During and After the Cold War. Journal of Politics 60(1): Neumayer, Eric and Laura Spess Do Bilateral Investment Treaties Increase Foreign Direct Investment to Developing Countries? World Development 33(10) Newnham, Randall Coalition of the Bribed and Bullied? U.S. Economic Linkage and the Iraq War Coalition. International Studies Perspective 9(2): Nielsen, Rich Matching with Time-Series Cross Sectional Data. Working Paper. Nielsen, Rich and Beth A. Simmons Rewards for Ratification: Payoffs for Participating in the International Human Rights Regime. International Studies Quarterly (forthcoming). Peinhardt, Clint and Todd Allee Failure to Deliver: The Investment Effects of Preferential Economic Agreements. World Economy 35(6): Poe, Steve C. and James Meernick US. Military Aid in the 1980s: A Global Analysis. Journal of Peace Research 32(4): Sachs, Lisa E. and Karl P. Sauvant BITs, DDTs, and FDI Flows: An Overview. In The Effect of Treaties on Foreign Direct Investment: Bilateral Investment Treaties, Double Taxation Treaties, and Investment Flows, edited by Karl P. Sauvant and Lisa E. Sachs, xxvii. New York, N.Y.: Oxford University Press. Sachs, Wayne The New U.S. Bilateral Investment Treaties. Berkley Journal of International Law 2(1):

41 Salacuse, Jeswald W BIT by BIT: The Growth of Bilateral Investment Treaties and Their Impact on Foreign Investment in Developing Countries. The International Lawyer 24(3): Salacuse, Jeswald W The Emerging Global Regime for Investment. Harvard International Law Journal 51(2): Salacuse, Jeswald W. and Nicholas P. Sullivan Do BITs Really Work? An Evaluation of Bilateral Investment Treaties Grand and Their Grand Bargain. Harvard International Law Journal 46(1): Shaffer, Gregory and Tom Gisnburg The Empirical Turn in International Legal Scholarship. American Journal of International Law 106(1):1-46. Simmons, Beth A Mobilizing for Human Rights: International Law in Domestic Politics. New York, N.Y.: Cambridge University Press. Simmons, Beth A. and Daniel J. Hopkins The Constraining power of International Treaties: Theory and Methods. American Political Science Review 99(4): Tobin, Jennifer L. and Marc L. Busch A Bit is Better than a Lot: Bilateral Investment Treaties and Preferential Trade Agreements. World Politics 62(1):1-42. Vandevelde, Kenneth J The BIT Program: A Fifteen-Year Appraisal. American Society of International Law Proceedings 86: Vandevelde, Kenneth J U.S. Bilateral Investment Treaties: The Second Wave. Michigan Journal of International Law14: Vandevelde, Kenneth J. 1998a. The Bilateral Investment Treaty Program of the United States. Cornell International Law Journal 21(2): Vandevelde, Kenneth J. 1998b. Investment Liberalization and Economic Development: The Role of Bilateral Investment Treaties. Columbia Journal of Transnational Law 36: Vandevelde, Kenneth J A Brief History of International investment Agreements. U.C. Davis Journal of International Law & Policy 12: Vandevelde, Kenneth J U.S. international Investment Agreements. New York, N.Y.: Oxford University Press. Verdier, Pierre-Hughes and Erik Voeten How Does Customary International Law CHane? The Case of State Immunity. Working Paper. Voeten, Erik Clashes in the Assembly. International Organization 54(2): Voeten, Erik Resisting the Lonely Superpower: Responses of States in the UN to U.S. Dominance. Journal of Politics 66(3):

42 von Stein, Jana Do Treaties Constrain or Screen? Selection Bias and Treaty Compliance. American Political Science Review 99: Yackee, Jason Webb Bilateral Investment Treaties, Credible Commitment, and the Rule of (International) Law: Do BITs Promote Foreign Direct Investment? Law and Society Review 42(4): Yackee, Jason Webb Do Bilateral Investment Treaties Promote Foreign Direct Investment? Some Hints from Alternative Evidence. Virginia Journal of International Law 51(2):

43 APPENDIXES Appendix A: Bilateral Investment Treaties Signed by the United States BIT Partner Signed Into Effect BIT Partner Signed Into Effect Panama 10/27/82 05/30/91 Ecuador 08/27/93 05/11/97 Senegal 12/06/83 10/25/90 Belarus 01/15/94 NA Haiti 12/13/83 NA Jamaica 02/04/94 03/07/97 D.R. Congo 08/03/84 07/28/89 Ukraine 03/04/94 11/16/96 Morocco 07/22/85 05/29/91 Georgia 03/07/94 08/17/97 Turkey 12/03/85 05/18/90 Estonia 04/19/94 02/16/97 Cameroon 02/26/86 04/06/89 Trinidad & Tobago 09/26/94 12/26/96 Egypt 03/11/86 06/27/92 Mongolia 10/06/94 01/01/97 Bangladesh 03/12/86 07/25/89 Uzbekistan 12/16/94 NA Grenada 05/02/86 03/03/89 Albania 01/11/95 01/04/98 Rep. Congo 02/12/90 08/13/94 Latvia 01/13/95 12/26/96 Poland 03/21/90 08/06/94 Honduras 07/01/95 07/11/01 Tunisia 05/15/90 02/07/93 Nicaragua 07/01/95 NA Sri Lanka 09/20/91 05/01/93 Croatia 07/13/96 06/20/01 Czech Republic 10/22/91 12/19/92 Jordan 07/02/97 06/12/03 Slovakia 10/22/91 12/19/92 Azerbaijan 08/01/97 08/02/01 Argentina 11/14/91 10/20/94 Lithuania 01/14/98 11/22/01 Kazakhstan 05/19/92 01/12/94 Bolivia 04/17/98 06/06/01 Romania 05/28/92 01/15/94 Mozambique 12/01/98 03/03/05 Russia 06/17/92 NA El Salvador 03/10/99 NA Armenia 09/23/92 03/29/96 Bahrain 09/29/99 05/30/01 Bulgaria 09/23/92 06/02/94 Uruguay 11/04/05 11/01/06 Kyrgyzstan 01/19/93 01/12/94 Rwanda 02/19/08 01/01/12 Moldova 04/21/93 11/25/94 - This table lists all BITs signed by the United States as of 12/31/ NA is used for BITs that have been signed but have not gone into effect. - an amended BIT was signed June 1, 2000 and went into effect May 14, Source: < (last visited 12/31/2013). 41

44 Appendix B: Senate Consideration of Bilateral Investment Treaties Country Introduced Passed Country Introduced Passed D.R. Congo 3/25/ /20/88 Jamaica 9/19/1994 6/27/96 Morocco 3/25/ /20/88 Belarus 9/23/1994 6/27/96 Senegal 3/25/ /20/88 Estonia 9/26/1994 6/27/96 Turkey 3/25/ /20/88 Ukraine 9/26/1994 6/27/96 Cameroon 5/28/ /20/88 Mongolia 6/26/1995 6/27/96 Bangladesh 5/30/ /20/88 Georgia 7/10/1995 6/27/96 Egypt 6/2/ /20/88 Latvia 7/10/1995 6/27/96 Grenada 6/3/ /20/88 Trinidad & Tobago 7/11/1995 6/27/96 Panama 3/25/ /28/90 Albania 11/6/1995 6/27/96 Poland 6/19/ /28/90 Uzbekistan 2/28/ /18/00 R. Congo 2/19/1991 8/11/92 Bahrain 5/23/ /18/00 Tunisia 5/17/1991 8/11/92 Bolivia 5/23/ /18/00 Sri Lanka 8/20/1991 8/11/92 Croatia 5/23/ /18/00 Czech Republic 6/2/1992 8/11/92 El Salvador 5/23/ /18/00 Slovakia 6/2/1992 8/11/92 Honduras 5/23/ /18/00 Russia 7/28/1982 8/11/92 Jordan 5/23/ /18/00 Kazakhstan 9/7/ /21/93 Mozambique 5/23/ /18/00 Romania 8/3/ /17/93 Lithuania 9/5/ /18/00 Argentina 1/19/ /17/93 Azerbaijan 9/12/ /18/00 Bulgaria 1/19/ /17/93 Uruguay 4/4/2006 9/12/06 Armenia 9/8/ /17/93 Rwanda 11/20/2008 9/26/11 Kyrgyzstan 9/8/ /17/93 Haiti 3/26/1986 NA Moldova 9/8/ /17/93 Nicaragua 6/26/2000 NA Ecuador 9/10/ /17/93 - This table lists all BITs signed by the United States as of 12/31/ NA is used for BITs that have been signed but have not gone into effect. - The date Introduced is the date that the treaty was introduced to the Senate. - The Passed date is the date that each BIT was approved by the Senate. - All BITs were passed by voice votes. 42

45 Appendix C: Correlation Matrix For Survival Analysis Investment Protection US FDI Outflow Military Aid Former Communist Polity Score GDP Per Capita Open Capital Accounts Prior BITs Investment Protection 1.00 US FDI Outflow Military Aid Former Communist Polity Score GDP Per Capita Open Capital Accounts Prior BITs